How a Title Sequence Sets the Tone Before the Story

Before the first line of dialogue, a good film has already told you what kind of night you are in for. A jagged arm over jazz, a spiral in an eye, a gun barrel tracking a man across the screen. Title designers spent the second half of the 20th century learning to set a tone with type and a minute of motion.

For a brand, on-screen type carries more than mood. It is the product name, the logo, the offer and the legal line, and it is what generated video does worst. AI video models paint letters as shapes, so a name can come out misspelled and a logo can shift between frames. The history of title design shows which part of that minute generation can handle, and which part it should never touch.

The arm that opened the curtain

In 1955, Otto Preminger’s The Man with the Golden Arm opened on white bars that appear, vanish and form patterns before settling into a distorted black arm, over Elmer Bernstein’s jazz score. Saul Bass, who designed it, wanted to convey “the disconnectedness and disjointedness of the addict’s life,” as Pat Kirkham records in her book on his work, excerpted by Art of the Title. He was after a spare, gaunt mood with a driving intensity.

Bernstein wrote his score while Bass worked on the design, so image and music grew together. And Preminger wanted the titles seen. Prints went out with a note telling projectionists to run the first reel only after the curtains had been drawn back. The opening was part of the film. Bass kept at it for four decades, from Hitchcock to Martin Scorsese’s Goodfellas and Casino.

Mood is the part of that job a model handles well. Abstract motion, light and texture contain nothing that must be spelled right, and a generation can try many backgrounds in an afternoon. Bass’s bars would make a fair prompt. The names riding on them would not: in a generated clip, every letter is redrawn on every frame.

Vertigo and Psycho: type that moves with the film

Bass opened Hitchcock’s Vertigo (1958) on extreme close-ups of Kim Novak’s face and spirals made with the animator John Whitney, a pioneer of computer art, who built them from mathematical curves known as Lissajous figures. Bernard Herrmann wrote the score. Writing for the BFI on Bass’s centenary, Matthew Thrift quotes Bass on the aim: to create “a climate for the story that was about to unfold.”

For Psycho (1960), Bass held himself to black, white and straight lines. Bars rush in from the edges of the frame to carry the sans-serif titles in and out, against Herrmann’s strings. Art of the Title’s Ben Radatz notes that they come from different parts of the screen yet never break formation or cross. The type itself fractures, and Thrift reads that fractured typography as the splintered mind of the film’s protagonist.

Psycho is a test generated video still fails. Testing ten text-to-video systems for T2VTextBench in 2025, researchers found that “most struggle to generate legible, consistent text.” A UCLA team reported invented symbols, smudged letters and text that changes from frame to frame. So the bars can be generated, while the words are set in a font and animated in compositing, timed to the bars.

A gun barrel, a pinhole and price stickers

Maurice Binder’s opening for Dr. No (1962) became one of the most repeated signatures in film. According to 007 Magazine, Binder could not get a photograph of a real gun barrel sharp until he “punched a pin-hole” in a piece of cardboard and held it to the lens. The gunshots that track across the screen began as white price stickers, and the red wash of blood was animated and combined with the footage in an optical printer. The man who turns and fires was the stunt arranger Bob Simmons, not Sean Connery, and the sequence first played in a trailer Binder assembled for UK cinemas. As Far Out Magazine points out, nothing yet suggested the series would become a major success.

The signature did not survive a change of format. The same article recounts that the original could not be adapted to widescreen Panavision anamorphic for Thunderball (1965), so a new version was shot with Connery himself.

A brand’s end card works like that barrel: the same logo, in the same proportions, every time. Each generation redraws a mark from what the model has learned, so letters swap and a wordmark can shimmer within one shot. Place the logo from its vector file in post and generate only the world around it. With type on its own layer, the Thunderball problem shrinks: a vertical cut becomes a new layout, rather than a crop that slices through the name.

Se7en and the killer’s handwriting

Kyle Cooper’s titles for David Fincher’s Se7en (1995), made at R/Greenberg Associates in Los Angeles, look like the opposite of Bass. Letters were etched by hand on scratchboard, notebooks full of text and images were filmed on a tabletop, and the result went through an optical printer. Cooper told Art of the Title that he mixed hand-drawn type with Helvetica because “it was from the mind of the killer.” The same article reports that the New York Times called the sequence one of the most important design innovations of the 1990s. In 2011, IFC ranked it third among the greatest title sequences ever made, and horror films in particular have imitated it ever since.

The mess was designed. Every scratch sits on letters someone chose and spelled.

That separates distressed type from a model’s garble: one reads as style, the other as a mistake. For a rough title, design the letters, then generate grain or scratches as a separate pass laid on top. For text inside the scene, such as a label or a sign, the VidScribe benchmark (September 2026) found across 11 systems that image-to-video kept existing text best, while editing text inside a generated video was the weakest task. Start from a real photo of the label, and check every frame.

The layers behind a title card in an AI project

Generation builds the bottom layer; everything with letters is set on top

The layers behind a title card in an AI video project

Layer	What it holds
1. Generated background	Mood, texture and motion, with no letters requested in the prompt
2. Real typography	Brand fonts and logo files, set and spaced by hand
3. Type animation	Movement of the type, timed to the background in compositing
4. Finishing pass	Grain, scratches or ink laid over clean letters
5. Delivery check	Type re-laid out for each aspect ratio and read on a phone at real size

What to ask of any AI video project

Ask which text in the video is generated and which is set in post. For logos, product names, prices and legal lines, the answer should be post, from the brand’s own files, with typefaces licensed for video. Ask to see every title on a phone at actual size: the W3C’s accessibility guidelines set a minimum contrast of 4.5:1 for normal text and 3:1 for large text, and they apply to text rendered into images. Ask how the square and vertical versions rebuild the type. Before approval, scrub any text inside the scene, on signs or packaging, frame by frame.